Ligandless Heck Coupling between a Halogenated Aniline and

May 16, 2008 - Peter Van der Donck,§ Tom Callewaert,§ Frank Meeussen,§ Erika De Bie,. ⊥ ... to develop a Heck coupling between the halogenated an...
0 downloads 0 Views 708KB Size
Organic Process Research & Development 2008, 12, 530–536

Ligandless Heck Coupling between a Halogenated Aniline and Acrylonitrile Catalyzed by Pd/C: Development and Optimization of an Industrial-Scale Heck Process for the Production of a Pharmaceutical Intermediate Didier Schils,*,†,2 Fred Stappers,† Geoffrey Solberghe,†,O Richard van Heck,† Michelle Coppens,† Dirk Van den Heuvel,‡ Peter Van der Donck,§ Tom Callewaert,§ Frank Meeussen,§ Erika De Bie,⊥ Kristof Eersels,¶ and Ellen Schouteden¶ Johnson and Johnson Pharmaceutical Research and DeVelopment, a DiVision of Janssen Pharmaceutica, API DeVelopment, Turnhoutseweg 30, B-2340 Beerse, Belgium

Abstract: The aniline derivative 3 is a key building block of rilpivirine (TMC278) 2, a new potent NNRTI compound under clinical evaluation. In this paper we describe the development of a new synthesis of 3 based on a Heck coupling between a halogenated aniline and acrylonitrile using low loading of Pd/C (0.5 mol %) as catalyst. This resulted in a process which has been successfully transferred into production on 2400 mol-scale (6000 L reactor)

Introduction Since its discovery in the early 70s by Mizoroki and Heck, the commonly called “Heck reaction” is one of the most powerful tools to create C-C bonds by reacting halogenated aromatics and olefins.1 In the past decade, new generations of homogeneous catalysts such as palladacycles, pincer-ligands, and more recently bulky electron-rich ligands have extended the scope of the reaction to all aromatic halides, even the poorly reactive aryl chlorides.2 Nevertheless, from the perspective of process development, homogeneous catalysis suffers from several drawbacks. First, ligands are required to form the active catalyst. These ligands may be expensive, not available in bulk or privileged. Also, the removal of the catalyst (the ligand, the metal or the complex) from the product can be challenging as the residual palladium at the API level has to be consistently under 10 ppm. For these reasons, only few Heck processes are used at production scale in the pharmaceutical industry.3 In this article, we would like to report the development of a Heck * Corresponding author. E-mail: [email protected]. † Chemical Process Research. ‡ Chemical Process Control. § Pilot Plant. ⊥ Safety Testing Laboratory. ¶ Chemical Production (Geel site, Belgium). 2 Current address: Euroscreen SA (47, Rue Adrienne Bolland, B-6041 Gosselies, Belgium). O Current address: UCB Pharma SA (Chemin du Foriest, B-1420 Braine-l’ Alleud, Belgium).

(1) (a) Mizoroki, T.; Mori, K.; Ozaki, A. Bull. Chem. Soc. Jpn. 1971, 44, 581. (b) Heck, R. F.; Nolley, J. P. J. Org. Chem. 1972, 37, 2320. (c) Review: Beletskaya, I. P.; Cheprakov, A. V. Chem. ReV. 2000, 100, 3009-3066 and references therein. (2) (a) Review: Littke, A.; Fu, G. Angew. Chem., Int. Ed. 2002, 41, 4176-4211. (b) Review: Whitecombe N. J.; Hii, K.; Gibson, S. Tetrahedron 2001, 57, 7449-7476. (c) Review oriented towards Process Chemistry: Farina, V. AdV. Synth. Catal. 2004, 346, 15531582. (3) (a) Beller, M.; Zapf, A.; Ma¨gerlein, W. Chem. Eng. Technol. 2001, 24, 575–582. (b) de Vries, J. G. Can. J. Chem. 2001, 79, 1086–1092. (c) Prashad, M. Top. Organomet. Chem. 2004, 6, 181–203. 530 • Vol. 12, No. 3, 2008 / Organic Process Research & Development Published on Web 05/16/2008

Figure 1. Structures of etravirine and rilpivirine Scheme 1

process which has been successfully transferred into production at full-scale (2400 mol/6000 L reactor). As part of our HIV program at Johnson & Johnson, two new non-nucleoside reversed transcriptase inhibitors (NNRTIs) are under active development: the recently FDA approved Intelence (etravirine, TMC125) 1 and rilpivirine (TMC278) 24 (Figure 1). For the commercial synthesis of TMC278, the intermediate 3 was required. This intermediate was first synthesized in a fourstep sequence from the commercially available 4-bromo-2,6dimethylaniline 44b and our goal became to develop a more efficient and cost-effective synthesis. Therefore we have decided to develop a Heck coupling between the halogenated aniline derivatives 4 or 5 and acrylonitrile that would provide 3 in a single step (Scheme 1). Results and Discussion Initial Screening. In order to gain an understanding of the reactivity of our aniline derivatives, the commercially available (4) (a) Janssen, P. A. J.; Lewi, P. J.; Arnold, E.; Daeyaert, F.; de Jonge, M.; Heeres, J.; Koymans, L.; Vinkers, M.; Guillemont, J.; Pasquier, E.; Kukla, M.; Ludovici, D.; Andries, K.; de Bethune, M.-P.; Pauwels, R.; Das, K.; Clark, A. D., Jr.; Frenkel, Y. V.; Hughes, S. H.; Medaer, B.; De Knaep, F.; Bohets, H.; De Clerck, F.; Lampo, A.; Williams, P.; Stoffels, P. J. Med. Chem. 2005, 48, 1901–1909. (b) Medicinal Chemistry synthesis of 3, unpublished results. 10.1021/op8000383 CCC: $40.75  2008 American Chemical Society

Table 1. Screening Heck reaction conditions of bromo- and iodo-derivatives with acrylonitrile entry

R

1a 1b 1c 1d

H

2a 2b 2c 2d

NO2

3a 3b 3c 3d

OMe

4a 4b 4c 4d

NH2

5a 5b 5c 5d

NH2

R1

X

cond.

yield (E + Z)a

E/Za

H

Br

A B A B

5.6 72.9 92.8 99.0

74/26 72/28 78/22 76/24

A B A B

100 100 91.8 100

74/26 75/25 80/20 75/25

A B A B

4.1 75.2 96.8 99.9

73/27 75/25 77/23 75/25

A B A B

6.5 29.9b 93.6 99.7

76/24 76/24 78/22 79/21

I H

Br I

H

Br I

H

Br I

Me

Br I

A B A B

3.0 87.5 100 94.3

80/20 80/20 79/21 80/20

a Yield (%) of products (E+Z isomers) from GLC analysis using an internal standard (diethylene glycol di-n-butyl ether). b Yield was 40% after 4 h but decreased with time as the Heck product reacts with acrylonitrile to give the corresponding Michael adduct (detected by GLC-MS).

Interestingly, results obtained with bromo- and iodo-derivatives 4 and 5 (entries 5a-d) showed that both halides can be used for the synthesis of the desired intermediate 3. For these two anilines, we have not observed any Michael addition of the aniline with acrylonitrile (compared to entry 4b). Presumably the two flanking methyl groups are sterically impeding this competing reaction. On the basis of these preliminary results, it was decided to develop processes with 4 and 5 in parallel in order to evaluate which one would be most suitable to be transferred to chemical production. Conditions for a Heck Reaction with 4-Iodo-2,6-dimethylaniline (5). Catalyst Loading. Palladium on charcoal (10% Pd/C, wet) is sufficient to catalyze the reaction, no supplementary ligand or salt being required (Table 1, entry 5c). The reaction is very selective, the main side products 6 (from dehalogenation) and 7 (double Heck product) were each formed in less than 1%, and we have found that it was possible to reach more than 99% conversion by using only 0.1 mol % of catalyst, although we have typically used loadings of 0.5 mol % to ensure robustness (Scheme 3, Table 2). Scheme 3

bromo derivative 4 and the easily synthesized iodo derivative 55 were compared as potential starting materials. Both homogeneous and heterogeneous Heck conditions6 were evaluated (Scheme 2, Table 1). Scheme 2 a

Table 2. Heck reaction of 3 with different catalyst loading (10% Pd/C, wet) entry

cat. loading (mol %)

3a

5a

Σ restb

1 2 3 4 5

0.5 0.4 0.2 0.1 0.05

96.2 94.2 91.5 96.3 84.0

0.4 0.7 12.9

3.4 5.8 8.5 3 3.1

a A: 10% Pd/C wet (0.05 equiv), DMA, NaOAc, 140°C, 20 h. B: Pd(OAc) 2 (0.05 equiv), tri-o-tolylphosphine (0.10 equiv), DMA, NaOAc, 140°C, 20 h.

a Determined by HPLC analysis (area % at 254 nm). b Sum of the total remaining impurities determined by HPLC analysis (area % at 254 nm).

As previously reported in the literature, all the Heck reactions with acrylonitrile afforded a mixture of E/Z-olefins in ratio ∼75/ 25 to 80/20. The iodides and the activated bromide reacted quantitatively using Pd/C as catalyst (entries 1 to 5/c,d and 2a,b). The nonactivated bromides gave low conversion (3-6.5%) with Pd/C (entries 1,3,4,5/a) but high conversion (75-87%) with the homogeneous system Pd(OAc)2/tri-o-tolylphosphine (entries 1,3,4,5/b).

Reaction Safety. The initial scale-up of this reaction revealed significant exothermic behavior. When all the compounds were mixed together and the mixture heated to 140 °C, the temperature rose from 110 to 150 °C in one minute. A common way to manage the exothermicity is to operate via a controlled addition of the reagents. RC1 experiment showed that addition of a solution of 5 and acrylonitrile in DMA (over 2 h) onto a warm suspension (140 °C) of Pd/C, NaOAc in DMA effectively controlled the exothermicity and allowed complete conversion after 17 h at 140 °C (Table 3).

(5) Kajigaeshi, S.; Kakinami, T.; Yamasaki, H.; Fujisaki, S.; Okamoto, T. Bull. Chem. Soc. Jpn. 1988, 61, 600. (6) (a) Homogeneous conditions adapted from: Herrmann, W.; Brossmer, C.; Ofele, K.; Reisinger, C. P.; Priermeier, T.; Beller, M.; Fischer, H. Angew. Chem., Int. Ed. Engl. 1995, 34, 1844-1847. (b) Heterogeneous conditions adapted from: Ko¨hler, K.; Heidenreich, R.; Krauter, J.; Pietsch, J. Chem. Eur. J. 2002, 8, 622-631. (7) Tr: reaction mixture temperature, Tj: jacket temperature, Qr: heat of reaction, Qb: baseline for Qr. The reaction enthalpy (∆Hr) is calculated by integration as the area between the Qr curve and the baseline Qb (∆Hr ) (Qr – Qb) · dt). The thermal conversion (R(t)) is calculated from the integral of Qr and normalized to 100%. Thermal conversion at the time (t) is given by R(t) ) ∆Hr(t)/∆Hr(E) x 100 where ∆Hr(t) ) reaction enthalpy from start to time t and ∆Hr(E) ) reaction enthalpy over the entire reaction time (In this case, initial time is dosing start).

Table 3. Conversion observed during the RC1 experiment (1.2 mol scale) timea (h)

3b

5b

Σ restc

3 6 17

76.2 91.3 96.2

21.8 5.7 0.4

2.0 2.8 3.4

a Including dosing time of 5. b Determined by HPLC analysis (area % at 254 nm). c Sum of the total remaining impurities determined by HPLC analysis (area % at 254 nm).

Vol. 12, No. 3, 2008 / Organic Process Research & Development



531

Figure 2. Temperature profile during the Heck reaction7 (RC1, 2-L glass reactor, Hastelloy Tr sensor, HC calibration probe, HC large propeller stirrer, and Ritter TG1 gas meter).

By using this mode of addition, the reaction can be safely performed at Pilot-Plant and Production scale. The reaction shows a medium exotherm (191 kJ/mol), and the accumulation at the end of the dosing is only 26% (Figure 2). Leaching and Residual Palladium. The amount of residual palladium at the level of intermediates 3 or 8 was assessed as part of our strategy for the control of the palladium content on the final API. We anticipated that the use of a heterogeneous “pre-catalyst” would be advantageous as the reactive palladium nanoparticles8 would agglomerate back onto the support surface upon completion of the reaction, allowing Pd removal by filtration of the catalyst.9 This assumption was confirmed by our results (Scheme 4). The leaching reached ∼27% (1340 ppm palladium in solution at 75% conversion) but decreased to only 10% (500 ppm) in the DMA solution after removal of the catalyst by filtration. The level of palladium can be further reduced by a simple treatment of the organic layer with activated charcoal (20 ppm of Pd remained in 3) which will ensure less than 5 ppm on the isolated product 8. Conditions for a Heck Reaction with 4-Bromo-2,6dimethylaniline (4). Catalyst Screening. A ligand screening for the bromoaniline derivative 4 using 5 mol % Pd (Pd/L: 1/2), DMA, NaOAc, 140 °C for 24 h showed that two phosphines, tri-o-tolylphosphine and tri-tert-butylphosphine, provided more than 90% conversion (Figure 3). We chose these two phosphines for further development. From the pioneer work of Jeffery on the influence of tetraalkylammonium salts on the Heck-type reaction, it is known that these salts can enhance the reactivity as well as the stability (8) For palladacycles, pincers and several heterogeneous Pd catalysts, the Heck coupling occurs at high temperature (120-160 °C). At these temperatures, they have tendency to form soluble Pd0 nanoparticles which are the true catalytic species. For more detailed discussion, see de Vries, J. G. Dalton Trans. 2006, 421–429 and references therein. (9) Review see: (a) Ko¨hler, K.; Pro¨ckl, S.; Kleist, W. Current Org. Chem. 2006, 10, 1585–1601, Review: (b) Phan, N.; Van Der Sluys, M.; Jones, C. AdV. Synth. Catal. 2006, 348, 609–679. (c) Thathagar, M.; ten Elshof, J.; Rothenberg, G. Angew. Chem., Int. Ed. 2006, 45, 2886– 2890. 532



Vol. 12, No. 3, 2008 / Organic Process Research & Development

Figure 3. Ligand screening for the Heck coupling (conversion based on HPLC results - area %). Scheme 4

Table 4. Screening Heck reaction conditions using Pd(OAc)2, P(o-tol)3 or P(t-Bu)3.HBF4, Bu4N+ClPd(OAc)2 P(o-tol)3 P(t-Bu)3 · Bu4NCla time conv.b entry (mol %) (mol %) HBF4(mol %) (equiv) (h) (area %) 1 2 3 4 5 6 7 8 9 10 11 b

5 5 4 3 2 1 1 5 1 0.5 0.25

10 10 8 6 4 2 2 – – – –

– – – – – – – 10 2 1 0.5

1 1.5 1 1 1 1 0.5 1 1 1 1

24 24 48 48 48 24 24 4 48 48 48

94.2 91.5 99.7 94.9 91.0 77.8 67.7 99.2 94.1 76.6 73.2

a Tetrabutylammonium chloride used is 95% purity (technical grade). determined by HPLC analysis (area % at 254 nm).

ligand, the following system was chosen for scale-up in the Pilot Plant: Pd/C (2.5 mol %)/P(o-tol)3 (5 mol %)/Bu4N+Cl(1 equiv) in DMA at 140 °C with NaOAc as base. Reaction Safety. The reaction with 4 is much less exothermic than the reaction with the iodoaniline 5 derivative and can be safely performed in batch mode in the Pilot Plant. This was confirmed by RC1 experiment where complete conversion was obtained after heating the reaction mixture for 20 h at 140 °C. Heck Processes with Bromo- and Iodo-Derivatives 4 and 5. Work-Up and Isolation. Isolation of the reaction product is an important part of a good process. For both reactions (with 4 or 5), removal of the catalyst by filtration, aqueous workup followed by a salt formation allowed us to isolate the hydrochloride salt 8 in 60-70% yield and high purity (HPLC > 98%) (Scheme 5).

of the catalyst.10 For the reaction of 4 with acrylonitrile (NaOAc, DMA, 140 °C), it was found that the use of Bu4N+Cl– 11 was beneficial with both phosphines (Table 4). With one equivalent of Bu4N+Cl-, complete conversion was obtained (entries 1,3,8,9) while using 1.5 equiv (entry 2) or 0.5 equiv (entry 7) showed reduced efficiency. Regarding the residual palladium content,12 we considered the use of a heterogeneous source of palladium in combination with a phosphine ligand. This concept has been extensively investigated by Lipshutz for Ni/C with external phosphine ligands and more recently by Merck researchers for Suzuki coupling with Pd/C13 Not surprisingly, we obtained similar results in the Heck coupling (Table 5).

Scheme 5

Table 5. Screening Heck reaction conditions using Pd/C (10% Pd/C, wet), P(o-tol)3 or P(t-Bu)3.HBF4 and Bu4N+Cl-

It is worth noting that during the salt formation, the E/Z ratio can be increased from 80/20 to 98/2. This is not due to an isomerization but rather to the selective destruction of the Z-isomer by Michael reaction with the solvent (Michael adduct is soluble in ethanol) and also to the higher solubility of the Zisomer in ethanol. Pilot-Plant Campaigns. These processes were implemented in our Pilot Plant, and several campaigns afforded more than 400 kg of 8 (250 kg from 5 and 150 kg from 4). Nevertheless, an issue in the isolation protocol was identified during the scaleup. In the laboratory filtration proceeded well mainly due to the fact that 8 forms agglomerates. However in the Pilot Plant, pressure filtration was changed to centrifugation, and the agglomerates were broken into very fine particles (8 µm) which obstructed the centrifugation bag, resulting in slower centrifugation and a less efficient washing effect.14 For the Heck process using iodo derivative 5, the residual palladium content was well under control (average 30 ppm with a maximum of 58 ppm). On the other hand, the bromo derivative 4 gave much higher residual Pd content with one case above 1000 ppm (96 to 1196 ppm) necessitating a rework.15 The extreme value (1196 ppm) is the result of bad centrifugation

conv.b Pd/C P(o-tol)3 P(t-Bu)3 · HBF4 Bu4NCla time (area Entry (mol %) (mol %) (mol %) (equiv) (h) %) 1 2 3 4 5 6 7 8

2.5 2.5 2.5 2.5 1.25 0.5 5 0.5

5 5 5 1 2.5 1 – –

– – – – – – 10 1

1 1c 0.5 1 1 1 1 1

20 4 24 48 24 24 24 24

99 98 87d 86 96.4 9.4 99 88

a Unless specified, tetrabutylammonium chloride used is 95% purity (technical grade). b determined by HPLC analysis (area % at 254 nm). c 98% purity tetrabutylammonium chloride (not usable in Plant due to high cost and low bulk availability).The reaction is faster because the technical grade tetrabutylammonium chloride contains up to 5% of tetrabutylammonium which is less reactive and slows down the reaction (see ref 11). d 12.5% (area %) double Heck product 7 observed.

On the basis of results of table 4 and 5, and taking into consideration the availability and the cost of the additive and (10) Review Jeffery, T. Tetrahedron 1996, 52, 10113–10130. (11) Other tetrabutylammonium salts were also tested and the reactivity is the following: AcO- ≈ F- > Cl- > Br- > HSO4- . I-. Lithium chloride gives 50% conversion while ammonium fluoride gives less than 10% conversion. For economical reasons, we selected the tetrabutylammonium chloride salt (technical grade 95% purity). (12) First results with homogeneous system (10 mol % Pd used) afforded 8 containing 6600 ppm of residual Pd. (13) (a) Lipshutz, B.; Blomgren, P. J. Am. Chem. Soc. 1999, 121, 5819– 5820. (b) Lipshutz, B.; Tasler, S.; Chrisman, W.; Spliethoff, B.; Tesche, B. J. Org. Chem. 2003, 68, 1177–1189. (c) Conlon, D.; Pipik, B.; Ferdinand, S.; LeBlond, C.; Sowa, J.; Izzo, B.; Collins, P.; Ho, G.; Williams, J.; Shi, Y.; Sun, Y. AdV. Synth. Catal. 2003, 345, 931–935.

(14) Interestingly these fine particles (8 µm) were observed only when the salt formation started from an 80/20 E/Z mixture (3). If the salt formation is performed on the pure E-isomer (E > 98%), much bigger crystals are obtained (average 250 µm), as determined by Lasentec experiments. All our efforts to improve the particle size of 8 starting from the 80/20 E/Z mixture 3 were unsuccessful. (15) Rework consisted of neutralization of the hydrochloride salt to liberate the base, extraction of the free base in toluene, treatment of the toluene phase with Silica-thiol followed by a solvent switch to ethanol and salt formation (yield: 80%). Vol. 12, No. 3, 2008 / Organic Process Research & Development



533

and poor washing effect. So for the bromo derivative 4 significant work would be needed for successful and robust transfer to production. Selection of the Final Process. Comparison of the two processes is given in Table 6.

to 80%.18 The final process was first scaled-up in the Pilot then successfully transferred in production at full scale in 6000 L reactors (Scheme 6). Scheme 6

Table 6. Comparison of Heck processes using the bromo-aniline 4 or the iodo-aniline 5 parameter

iodoaniline process

yield purity max. residual Pd processability process Cost cost of aniline cost of catalytic system potential for cost decrease

same (∼65%) same (>98%) 58 ppm same same high low high

bromoaniline process

1196 ppm low high low

Both processes have comparable yield, quality, processability, and cost. However two parameters are in favor of iodoaniline process 5. First this process was deemed more robust allowing easy and reliable control of the residual palladium content. Second, another advantage of the iodoaniline process is its potential for cost decrease. Although both processes are equivalent in cost, the cost drivers are different. For the iodoaniline process, the major part of the cost is the iodoaniline 5 (80% of the total process cost). For the bromoaniline process, the cheap bromoaniline 4 represents only 20% of the cost, while the catalytic system, which will remain constant, is contributing for more than 70% (higher Pd loading, phosphine ligand, and ammonium salt required). This means that any price decrease of 5 (which is expected with increasing scale) will have a high impact of the total process cost, rendering the iodoaniline process more attractive.16 Our example illustrates the fact that it is sometimes easier and cheaper to adapt the substrate (choice of iodoaniline 5) to an inexpensive catalyst (Pd/C) rather than to spend significant resources to find a catalytic system able to perform the same transformation on a cheap and commercially available aryl bromide (or chloride). For these reasons, the iodoaniline 5 was selected for the final synthesis of rilpivirine. Final Process Implemented in Chemical Production. In addition to the benefits mentioned above, the iodoaniline 5 afforded pure enough 3 (low Pd level) that we could use in the next step of the synthesis.17 This allowed us to avoid the difficult isolation of 8 while increasing the overall yield of the reaction (16) As mentioned before in this article, at the beginning of the project, the 4-iodo-2,6-dimethylaniline 5 was not commercially available. After one year, two suppliers were found. Until now, more than seven suppliers have been evaluated. At the time we had to select the process, the price of 5 had already decreased by more than 15 times since our first order. Since then, significant decrease occurred again, confirming our assumption made about the potential of cost decrease. The iodoaniline process (cost for production of 3) is now 25% cheaper than the bromoaniline process. (17) The starting E/Z ratio of the aniline (80/20 for 3 or 98/2 for 8) is of little importance since a thermodynamic ratio (E/Z: 85/15) is obtained in the reaction conditions of next step of the synthesis. 534



Vol. 12, No. 3, 2008 / Organic Process Research & Development

We were delighted to notice that the exothermicity was completely under control, even when the reaction was performed at full scale in our production facilities. It is also important to mention that no emission of acrylonitrile has been detected during these campaigns.19 The results of the pilot plant and production campaigns show the robustness and the efficiency of the process. In four production batches more than 1.3 ton of 3 has been produced (Table 7). Table 7. Overview of Pilot Plant batches (260 mol) and Chemical Production batches (2388 mol) entry

mol-scale

conc.a

purityb

yieldc

residual Pdd

1 2 3 4 5 6 7 8

260 260 260 260 2388 2388 2388 2388

21.9 21.9 21.5 21.3 21.1 20.9 20.5 21.3

93.7 94.6 93.0 93.4 94.6 95.3 94.7 94.7

75.7 81.8 79.3 80.9 82 81 80 82

56 40 28 35